Re: RE : Re: RE : ntl
On Wed, Aug 10, 2011 at 12:54 PM, Bastien ROUCARIES
<roucaries.bastien@gmail.com> wrote:
> Ok in this case use the wrapper from gnulib
>
> Bastien
>
> Le 10 août 2011 12:04, "Bernhard R. Link" <brlink@debian.org> a écrit :
>
> * roucaries bastien <roucaries.bastien+debian@gmail.com> [110810 00:09]:
>
>> On Tue, Aug 9, 2011 at 5:03 PM, Bernhard R. Link <brlink@debian.org>
>> wrote:
>
>> > I think the behaviour ntl is using (some assembler on i386 to unset and
>> > restore the CPU bit ...
>
> Which other lib would want to do floating point operations between the
> start and the end of one of the functions of quad_float.c ?
Did you read this http://gcc.gnu.org/onlinedocs/gcc/Floating-Types.html
A little bit out of topic but worth to report upstream
After reviewing the implementation it seems that it is quite fragile:
1. #if (defined(__GNUC__) && NTL_EXT_DOUBLE && (defined(__i386__) ||
defined(__i486__) || defined(__i586__)))
Does not work if 686 is compiled with -mfpmath=387
2. volatile trick is fragile
Care to report upstream ?
Bastien
> Bernhard R. Link
>
>
> --
> To UNSUBSCRIBE, email to debian-science-request@lists.debian.org
> with a sub...
>
> Archive:
> [🔎] 20110810100530.GA18450@pcpool00.mathematik.uni-freiburg.de">http://lists.debian.org/[🔎] 20110810100530.GA18450@pcpool00.mathematik.uni-freiburg.de
>
>
Reply to: